Cross-linguistic studies of aphasia permit us to separate universal mechanisms from language-specific content. By uncovering the range of variations that are possible under normal and abnormal conditions, cross-language studies also address the critical issues of behavioral and neural plasticity. In this proposal, we outline new comparative studies of language processing and language breakdown in aphasic patients and controls in three languages (English, Italian and Chinese) that differ dramatically in their lexical and grammatical structure (e.g. amounts of word order variation, inflectional morphology, constituent omission, consistency vs. irregularity of words and morphemes, potential for lexical ambiguity, and the internal structure of words). Patient studies (the classical method of lesion-behavior mapping) are complemented by brain-imaging studies of normals in the same three languages (using functional magnetic resonance imaging, or fMRI). The same materials are used in behavioral and fMRI experiments, in 'on-line', computer- controlled tasks that yield information about the temporal dynamics of word and sentence processing. Nonverbal control tasks are designed to match linguistic tasks in key respects (visual, auditory, and motor activation; demands on memory, attention, decision-making), testing hypotheses about the contributions of modality and sensorimotor demands to language activation (fMRI) and language breakdown (lesion studies). We also expand the concept of "normal control" to include comparisons of normal tested under adverse processing conditions (perceptual degradation, temporal compression, cognitive overload), to uncover 'breakpoints' in processing and to 'simulate' processing disorders in patients. Selection of word and picture stimuli is based on massive norming information collected at all sites in the last funding cycle. The aphasia subgroups under study include nonfluent Broca's aphasics, fluent Wernicke's aphasics, and anomic patients who commit few overt grammatical errors but still struggle to 'find the right word'. Acknowledging the limitations of traditional aphasia categories, we also take a new multivariate approach, analyzing patients' performance on experiments within a continuous, multidimensional symptom space, defined for each language by using large archival data sets (more than 200 patients per language). Results are interpreted within a merger of two theoretical frameworks: the Competition Model (a processing model that assumes interactive activation over distributed, probabilistic representations) and Embodiment Theory (a theory of neural organization for language).